CN115775022A - An optical pulse neural network processor device - Google Patents

Abstract

An optical pulse neural network processor device comprises a single-chip integrated and packaged SOA and DFB-SA array optical pulse neural network chip, a multi-channel tunable light source, a modulator array, a photoelectric detector PD array, a transimpedance amplifier array, an analog-to-digital conversion array, two digital-to-analog conversion arrays, a control circuit array and a memory array. The optical pulse neural network chip uses a semiconductor optical amplifier SOA as a photon synapse, and uses a distributed feedback semiconductor laser DFB-SA with a saturated absorption region as a photon neuron. The photon synapse and the photon neuron are both based on III/V materials, so the sub-photon pulse neural network chip can be monolithically integrated. The device ensures that the optical pulse neural network chip can be integrated on a single chip and the optical pulse neural network processor is flexibly configurable.

Description

An optical pulse neural network processor device

technical field

A device for an optical pulse neural network processor belongs to the field of optical calculation and optical signal processing, and specifically relates to a device for an optical pulse neural network processor based on a semiconductor optical amplifier and a distributed feedback laser with a saturated absorption region. It is characterized in that its optical pulse neural network chip is based on III/V group monolithic integration.

Background technique

Compared with traditional von Neumann computing systems, spiking neural network computing seeks to draw inspiration from the abstract principles of biological information processing to design scalable and cost-effective hardware systems. In tasks involving pattern analysis, decision-making, optimization, learning, and real-time control of multi-sensor, multi-actuator systems, neuromorphic architectures have powerful advantages over von Neumann architectures in terms of efficiency, fault tolerance, and adaptability.

The vigorous development of optical integration technology has greatly promoted high-performance optical computing to become the world's frontier hot spot again. The on-chip optical pulse neural network fully integrates the characteristics of high-speed optical communication, optical interconnection, optical integration and neuromorphic computing, and has the advantages of ultra-high speed, large bandwidth, and multi-dimensionality. It has broad application prospects in the fields of high-performance computing and artificial intelligence.

The basic functional unit of photon spiking neural network is photon spiking neuron and photon synapse. At present, some progress has been made in photonic pulse neurons and optical synapses based on discrete devices, or optical pulse neural networks based on heterogeneous integration. However, the optical pulse neural network for monolithic integration is still in the cutting-edge exploration stage. Therefore, it is urgent to develop a monolithically integrated optical pulse neural network chip and a processor based on the optical pulse neural network chip.

Contents of the invention

In view of the deficiencies of the prior art stated above, the present invention aims to provide an optical pulse neural network processor device. The optical pulse neural network chip in the device uses a semiconductor optical amplifier SOA as a photon synapse, and a distributed feedback semiconductor laser DFB-SA with a saturated absorption region as a photon pulse neuron. Both the photonic synapse and the photonic pulse neuron are based on III/V materials, so the photonic pulse neural network chip can be monolithically integrated.

The purpose of the present invention is achieved by the following means.

A device for an optical pulse neural network processor, comprising a single-chip integrated and packaged SOA and DFB-SA array optical pulse neural network chip, a multi-channel tunable light source, a modulator array, a photodetector PD array, A transimpedance amplifier array, an analog-to-digital conversion array, two digital-to-analog conversion arrays, a control circuit array and a memory array. A device for an optical pulse neural network processor is characterized in that its optical pulse neural network chip is based on III/V monolithic integration, and the output end of the multi-channel tunable light source is connected to the optical input end of the modulator array ; The output ends of the modulator array are respectively connected to the input ends of the optical pulse neural network chip; the two output ends of the controller are connected to the input ends of the two digital-to-analog conversion arrays; The RF end of the device is connected to the current control end of the optical pulse neural network chip; the output end of the optical pulse neural network chip is connected to the input end of the PD array; the output end of the PD array is connected to the input end of the transimpedance amplifier array; the transimpedance amplifier array The output end of the analog-to-digital conversion array is connected to the input end of the analog-to-digital conversion array; the output end of the analog-to-digital conversion array is connected to the input end of the control circuit; the control circuit and the memory are connected to each other for data storage and reading. Regarding the optical pulse neural network chip, it is characterized in that the output end of the modulator is connected with the input end of the on-chip beam splitter; the output end of the beam splitter is connected with the input end of the arrayed waveguide grating; the output end of the arrayed waveguide grating is connected with the SOA connected to the input of the synapse; the output of the SOA synapse is connected to the input of the second arrayed waveguide grating; the output of the second arrayed waveguide grating is connected to the DFB-SA neuron array; the output of the DFB-SA neuron array As the output of the optical pulse neural network chip, it is transformed and analyzed in the next step.

After the above design, single-channel SOA and DFB-SA and other discrete devices are used as examples for testing; the injection pulse of the optical pulse neural network is composed of tunable light source TL, polarization controller PC, modulator, arbitrary waveform transmitter AWG and other devices Generated; the external light injection is weighted through the SOA synapse; then it enters the DFB-SA through the adjustable optical attenuator VOA and the three-port circulator CIR; the signal after the action of the DFB-SA is output to the spectrum analyzer through the CIR and the optical coupler OC OSA with oscilloscope OSC; the output of DFB-SA neurons was observed by adjusting the current of SOA synapses to make SOA produce different weights.

Compared with the reported optical pulse neural network processor device, the device of the present invention provides an optical pulse neural network processor has the following advantages: the optical pulse neural network chip can be single-chip integrated.

Description of drawings

Fig. 1 is the system block diagram of device of the present invention;

Fig. 2 is a schematic diagram of the internal structure of the optical pulse neural network chip;

Figure 3 is a diagram of a single-channel SOA, DFB-SA optical pulse neural network experiment scheme;

Figure 4 shows the threshold characteristics of the optical pulse neural network to external stimuli

Figure 5 is the cumulative response of the light pulse neural network to external stimuli.

Figure 6 shows the refractory period response of the light pulse neural network to external stimuli.

Detailed ways

Below in conjunction with accompanying drawing the embodiment of the present invention is described in detail: present embodiment is carried out under the premise of technical solution of the present invention, has provided detailed implementation mode and concrete operation process, but protection scope of the present invention is not limited to subordinates the embodiment.

As shown in Figure 3, the embodiment of the present invention consists of a tunable light source TL, two polarization controllers PC1, PC2, an intensity modulator, an electrical amplifier, an AWG, two dual-channel DC sources, and a semiconductor optical amplifier SOA, two adjustable optical attenuators VOA1, VOA2, a three-port optical circulator CIR, an optical coupler OC, a DFB-SA, a current temperature controller, a photodetector PD, oscilloscope OSC, spectrometer OSA . The SOA is a photon synapse; DFB-SA is a photon pulse neuron; SOA and DFB-SA form a photon pulse neural network; by adjusting the current of SOA, that is, the synapse weight in the photon pulse neural network, photon pulses can be observed Differential responses of spiking neural networks to stimuli.

In this example, the specific implementation steps of the method are:

Step 1: In the solution shown in Figure 3, the DFB-SA is a two-stage semiconductor laser with a wavelength of 1550-1554nm. Set the wavelength of TL to 1552.3nm.

Step 2: Adjust the current and bias voltage of the gain region and the saturated absorption region of the DFB-SA to 32mA and -2.8V respectively. Set the AWG so that one large and one small stimulation pulse enters the optical pulse neural network. The stimulation pulse is shown in Figure 4(a) to adjust the SOA synaptic current and observe the dynamic response of DFB-SA. The result is shown in Figure 4(b) ) shown. It can be seen from Figure 4(b) that under the action of SOA, DFB-SA has no impulse response under small-intensity stimulation, but has an impulse response under high-intensity stimulation, realizing the threshold characteristic of the optical impulse neural network.

Step 3: Set the AWG so that three stimulation pulses with small intervals enter the optical pulse neural network, and the stimulation pulses are shown in Figure 5(a). Adjust the SOA synaptic current and observe the dynamic response of DFB-SA, the results are shown in Figure 5(b). It can be seen from Figure 5(b) that under the action of SOA, the DFB-SA photon pulse neuron only produces a spike at the third stimulation pulse, which shows that the optical pulse neural network can produce a cumulative effect.

Step 4: Adjust the SOA synaptic current again and observe the dynamic response of DFB-SA. The neurons only produced spikes at the first and third stimulation pulses, indicating that the light-spiked neural network can produce refractory periods.

Step 5: Increase the number of SOA synapses and DFB-SA neurons of the optical pulse neural network, and form a multi-node optical pulse neural network chip through integration, as shown in Figure 2.

Step 6: Add electro-optic interface, photoelectric interface, control circuit, etc. on the basis of Figure 2 to form a configurable optical pulse neural network processor, as shown in Figure 1.

Based on the above statements, the present invention has the following characteristics: 1). Utilize a semiconductor optical amplifier array as a synapse array, and a distributed feedback semiconductor laser array with a saturated absorption zone as a neuron array to form a monolithically integrated optical pulse neural network chip; 2). Based on the optical pulse neural network chip, electro-optical/optical interface, control circuit and other modules, a configurable optical pulse neural network processor is formed.

In a word, the embodiments described above are only examples of the present invention, and are not only used to limit the protection scope of the present invention. Several equivalent deformations and replacements (such as appropriately changing the size of the operating current, changing the frequency detuning, changing the size of the injected pulse power, and expanding the scale of the optical pulse neural network) should also be included within the protection scope of the present invention.

Claims (4)

1. An optical pulse neural network processor device comprises a monolithic integration and packaged SOA and DFB-SA array optical pulse neural network chip, a multi-channel tunable light source, a modulator array, a photoelectric detector PD array, a transimpedance amplifier array, an analog-to-digital conversion array, two digital-to-analog conversion arrays, a control circuit array and a memory array, and is characterized in that the optical pulse neural network chip is based on III/V family monolithic integration, and the output end of the multi-channel tunable light source is connected with the optical input end of the modulator array; the output end of the modulator array is respectively connected with the input end of the optical pulse neural network chip; two output ends of the controller are connected with input ends of the two digital-to-analog conversion arrays; the output ends of the two digital-to-analog conversion arrays are respectively connected with the radio frequency end of the modulator and the current control end of the optical pulse neural network chip; the output end of the optical pulse neural network chip is connected with the input end of the PD array; the output end of the PD array is connected with the input end of the transimpedance amplifier array; the output end of the transimpedance amplifier array is connected with the input end of the analog-to-digital conversion array; the output end of the analog-to-digital conversion array is connected with the input end of the control circuit; the control circuit is connected with the memory to store and read data. The optical pulse neural network chip is characterized in that the output end of the modulator is connected with the input end of the on-chip beam splitter; the output end of the beam splitter is connected with the input end of the array waveguide grating; the output end of the array waveguide grating is connected with the input end of the SOA synapse; the output end of the SOA synapse is connected with the input end of the second arrayed waveguide grating; the output end of the second arrayed waveguide grating is connected with the DFB-SA neuron array; and the output of the DFB-SA neuron array is used as the output of the optical pulse neural network chip for further transformation and analysis.

2. The optical pulse neural network processor device as claimed in claim 1, wherein a Semiconductor Optical Amplifier (SOA) is used as a photonic synapse, and a feedback semiconductor laser (DFB-SA) with a saturated absorption region is used as a photonic neuron; the photonic synapses and the photonic neurons are based on III/V materials, and the optical pulse neural network chip can be monolithically integrated.

3. The apparatus of claim 1, wherein the optical pulse neural network processor is flexibly configurable via connection of an optoelectronic/optoelectronic interface, a control circuit, and the like.

4. The apparatus of claim 1, wherein the number of nodes of the optical pulse neural network chip is expandable to n x n.

Images